MNIST_Handwritten_Digits_STARTER_(1).ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PL4Pj_Uf-UKX"
      },
      "source": [
        "# Introduction\n",
        "\n",
        "In this project, you will build a neural network of your own design to evaluate the MNIST dataset.\n",
        "\n",
        "Some of the benchmark results on MNIST include can be found [on Yann LeCun's page](https://webcache.googleusercontent.com/search?q=cache:stAVPik6onEJ:yann.lecun.com/exdb/mnist) and include:\n",
        "\n",
        "88% [Lecun et al., 1998](https://hal.science/hal-03926082/document)\n",
        "\n",
        "95.3% [Lecun et al., 1998](https://hal.science/hal-03926082v1/document)\n",
        "\n",
        "99.65% [Ciresan et al., 2011](http://people.idsia.ch/~juergen/ijcai2011.pdf)\n",
        "\n",
        "\n",
        "MNIST is a great dataset for sanity checking your models, since the accuracy levels achieved by large convolutional neural networks and small linear models are both quite high. This makes it important to be familiar with the data.\n",
        "\n",
        "## Installation"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "6MBZHtyR-UKj"
      },
      "outputs": [],
      "source": [
        "# Update the PATH to include the user installation directory.\n",
        "import os\n",
        "os.environ['PATH'] = f\"{os.environ['PATH']}:/root/.local/bin\"\n",
        "\n",
        "# Restart the Kernel before you move on to the next step.|"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gQj6kukx-UKn"
      },
      "source": [
        "#### Important: Restart the Kernel before you move on to the next step."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "o7MYNnvf-UKp"
      },
      "outputs": [],
      "source": [
        "# Install requirements\n",
        "!python -m pip install -r requirements.txt"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "m9VU3V4y-UKr"
      },
      "source": [
        "## Imports"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "LZSSaHbb-UKs"
      },
      "outputs": [],
      "source": [
        "## This cell contains the essential imports you will need – DO NOT CHANGE THE CONTENTS! ##\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "import torch.optim as optim\n",
        "import torchvision\n",
        "import torchvision.transforms as transforms\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "_zLGodXJAyTR"
      },
      "outputs": [],
      "source": [
        "from torch.utils.data import DataLoader\n",
        "from torchvision import transforms\n",
        "from torchvision import datasets\n",
        "from torchvision.transforms import ToTensor\n",
        "import matplotlib.pyplot as plt"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gF9250OTcgvZ",
        "outputId": "4efd9a6f-76da-4f8b-ef26-02ce60b1d7f6"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "device(type='cpu')"
            ]
          },
          "execution_count": 9,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
        "device"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AXDWEv5S-UKu"
      },
      "source": [
        "## Load the Dataset\n",
        "\n",
        "Specify your transforms as a list if you intend to .\n",
        "The transforms module is already loaded as `transforms`.\n",
        "\n",
        "MNIST is fortunately included in the torchvision module.\n",
        "Then, you can create your dataset using the `MNIST` object from `torchvision.datasets` ([the documentation is available here](https://pytorch.org/vision/stable/datasets.html#mnist)).\n",
        "Make sure to specify `download=True`!\n",
        "\n",
        "Once your dataset is created, you'll also need to define a `DataLoader` from the `torch.utils.data` module for both the train and the test set."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "5ZYjNlAt-UKw"
      },
      "outputs": [],
      "source": [
        "# Define transforms\n",
        "## YOUR CODE HERE ##\n",
        "# transform = transforms.Compose([transforms.ToTensor(),transforms.flatten()])\n",
        "transform = transforms.Compose([transforms.ToTensor(),transforms.Normalize((0.5,), (0.5,))])\n",
        "\n",
        "## YOUR CODE HERE ##\n",
        "training_data=datasets.MNIST(root=\"data\",train=True,download=True,transform=transform)\n",
        "train_loader=DataLoader(training_data,batch_size=32,shuffle=True)\n",
        "# Create test set and define test dataloader\n",
        "## YOUR CODE HERE ##\n",
        "testing_data=datasets.MNIST(root=\"data\",train=False,download=True,transform=transform)\n",
        "test_loader=DataLoader(testing_data,batch_size=32)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "V6ZcH4ql-UKz"
      },
      "source": [
        "## Justify your preprocessing\n",
        "\n",
        "In your own words, why did you choose the transforms you chose? If you didn't use any preprocessing steps, why not?"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qtM5p0le-UK0"
      },
      "source": [
        "**DOUBLE CLICK THIS CELL TO MODIFY**\n",
        "\n",
        "> first i chose toTensor to convert the data into multi dimensional arrays for faster computation for the images and then chose to normalise the data so that it becomes easier for computation since the data would lie in the same range and further on i flatten the image to make it available in the one dimension .\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fL6C4Rsu-UK1"
      },
      "source": [
        "## Explore the Dataset\n",
        "Using matplotlib, numpy, and torch, explore the dimensions of your data.\n",
        "\n",
        "You can view images using the `show5` function defined below – it takes a data loader as an argument.\n",
        "Remember that normalized images will look really weird to you! You may want to try changing your transforms to view images.\n",
        "Typically using no transforms other than `toTensor()` works well for viewing – but not as well for training your network.\n",
        "If `show5` doesn't work, go back and check your code for creating your data loaders and your training/test sets."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "beXVnjTb-UK2"
      },
      "outputs": [],
      "source": [
        "## This cell contains a function for showing 5 images from a dataloader – DO NOT CHANGE THE CONTENTS! ##\n",
        "def show5(img_loader):\n",
        "    dataiter = iter(img_loader)\n",
        "\n",
        "    batch = next(dataiter)\n",
        "    labels = batch[1][0:5]\n",
        "    images = batch[0][0:5]\n",
        "    for i in range(5):\n",
        "        print(int(labels[i].detach()))\n",
        "\n",
        "        image = images[i].numpy()\n",
        "        plt.imshow(image.T.squeeze().T)\n",
        "        plt.show()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "Y_Ftn3CY-UK4",
        "outputId": "18ad35ce-3af1-48f5-e7b9-f98dad6995ae"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "1\n"
          ]
        },
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "9\n"
          ]
        },
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "0\n"
          ]
        },
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "3\n"
          ]
        },
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "3\n"
          ]
        },
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Explore data\n",
        "## YOUR CODE HERE ##\n",
        "show5(train_loader)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zs-cptal-UK6"
      },
      "source": [
        "## Build your Neural Network\n",
        "Using the layers in `torch.nn` (which has been imported as `nn`) and the `torch.nn.functional` module (imported as `F`), construct a neural network based on the parameters of the dataset.\n",
        "Use any architecture you like.\n",
        "\n",
        "*Note*: If you did not flatten your tensors in your transforms or as part of your preprocessing and you are using only `Linear` layers, make sure to use the `Flatten` layer in your network!"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PKRnWmyr-UK7",
        "outputId": "caceab74-cab5-40a9-aa2d-bd8fa8c7daed"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "Net(\n",
              "  (fc1): Linear(in_features=2352, out_features=120, bias=True)\n",
              "  (fc2): Linear(in_features=120, out_features=84, bias=True)\n",
              "  (fc3): Linear(in_features=84, out_features=10, bias=True)\n",
              ")"
            ]
          },
          "execution_count": 37,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "## YOUR CODE HERE ##\n",
        "# Define the class for your neural network\n",
        "class MNIST_MLP(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(MNIST_MLP, self).__init__()\n",
        "        self.fc1 = nn.Linear(28 * 28, 120)  # Corrected input dimension for MNIST\n",
        "        self.fc2 = nn.Linear(120, 84)\n",
        "        self.fc3 = nn.Linear(84, 10)\n",
        "        self.activation = nn.ReLU()\n",
        "        self.output = nn.LogSoftmax(dim=1)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = torch.flatten(x, 1)  # Flatten the input tensor\n",
        "        x = self.activation(self.fc1(x))\n",
        "        x = self.activation(self.fc2(x))\n",
        "        x = self.output(self.fc3(x))\n",
        "        return x\n",
        "\n",
        "\n",
        "# Instantiate the model\n",
        "net = Net()\n",
        "net.to(device)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jgI4raB1-UK7"
      },
      "source": [
        "Specify a loss function and an optimizer, and instantiate the model.\n",
        "\n",
        "If you use a less common loss function, please note why you chose that loss function in a comment."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "NLnd5dZn-UK8"
      },
      "outputs": [],
      "source": [
        "## YOUR CODE HERE ##\n",
        "criterion = nn.NLLLoss()\n",
        "optimizer = optim.SGD(net.parameters(), lr=0.005, momentum=0.9)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WAFXCq4y-UK-"
      },
      "source": [
        "## Running your Neural Network\n",
        "Use whatever method you like to train your neural network, and ensure you record the average loss at each epoch.\n",
        "Don't forget to use `torch.device()` and the `.to()` method for both your model and your data if you are using GPU!\n",
        "\n",
        "If you want to print your loss **during** each epoch, you can use the `enumerate` function and print the loss after a set number of batches. 250 batches works well for most people!"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "3nZWQaTM-UK_"
      },
      "outputs": [],
      "source": [
        "mlp = MNIST_MLP()\n",
        "if torch.cuda.is_available():\n",
        "    mlp.cuda()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NmgsTT5e-ULB"
      },
      "source": [
        "Plot the training loss (and validation loss/accuracy, if recorded)."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 50,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 777
        },
        "id": "74pta10--ULC",
        "outputId": "a561c7a3-e25d-48eb-90ad-3e35143449dd"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1 training accuracy: 99.99% training loss: 0.00072\n",
            "Epoch 1 validation accuracy: 98.39% validation loss: 0.10507\n",
            "Epoch 2 training accuracy: 99.99% training loss: 0.00055\n",
            "Epoch 2 validation accuracy: 98.41% validation loss: 0.10533\n",
            "Epoch 3 training accuracy: 99.99% training loss: 0.00051\n",
            "Epoch 3 validation accuracy: 98.46% validation loss: 0.10756\n",
            "Epoch 4 training accuracy: 100.00% training loss: 0.00040\n",
            "Epoch 4 validation accuracy: 98.47% validation loss: 0.10920\n",
            "Epoch 5 training accuracy: 99.99% training loss: 0.00037\n",
            "Epoch 5 validation accuracy: 98.42% validation loss: 0.10847\n",
            "Epoch 6 training accuracy: 100.00% training loss: 0.00033\n",
            "Epoch 6 validation accuracy: 98.44% validation loss: 0.11020\n",
            "Epoch 7 training accuracy: 100.00% training loss: 0.00029\n",
            "Epoch 7 validation accuracy: 98.38% validation loss: 0.11063\n",
            "Epoch 8 training accuracy: 100.00% training loss: 0.00026\n",
            "Epoch 8 validation accuracy: 98.46% validation loss: 0.11228\n",
            "Epoch 9 training accuracy: 99.99% training loss: 0.00029\n",
            "Epoch 9 validation accuracy: 98.41% validation loss: 0.11284\n",
            "Epoch 10 training accuracy: 100.00% training loss: 0.00025\n",
            "Epoch 10 validation accuracy: 98.43% validation loss: 0.11346\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "## YOUR CODE HERE ##\n",
        "# Function to train the network\n",
        "def train_network_classification(net, train_loader, test_loader):\n",
        "    num_epochs = 10\n",
        "\n",
        "    criterion = nn.NLLLoss()\n",
        "    optimizer = optim.SGD(net.parameters(), lr=0.005, momentum=0.9)\n",
        "\n",
        "    train_loss_history = []\n",
        "    val_loss_history = []\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        net.train()\n",
        "        train_loss = 0.0\n",
        "        train_correct = 0\n",
        "        for i, data in enumerate(train_loader):\n",
        "            inputs, labels = data\n",
        "\n",
        "            if torch.cuda.is_available():\n",
        "                inputs, labels = inputs.cuda(), labels.cuda()\n",
        "\n",
        "            optimizer.zero_grad()\n",
        "            outputs = net(inputs)\n",
        "            loss = criterion(outputs, labels)\n",
        "            loss.backward()\n",
        "            optimizer.step()\n",
        "\n",
        "            _, preds = torch.max(outputs.data, 1)\n",
        "            train_correct += (preds == labels).sum().item()\n",
        "            train_loss += loss.item()\n",
        "        print(f'Epoch {epoch + 1} training accuracy: {train_correct/len(training_data)*100:.2f}% training loss: {train_loss/len(train_loader):.5f}')\n",
        "        train_loss_history.append(train_loss)\n",
        "\n",
        "        val_loss = 0.0\n",
        "        val_correct = 0\n",
        "        net.eval()\n",
        "        with torch.no_grad():\n",
        "            for inputs, labels in test_loader:\n",
        "                if torch.cuda.is_available():\n",
        "                    inputs, labels = inputs.cuda(), labels.cuda()\n",
        "\n",
        "                outputs = net(inputs)\n",
        "                loss = criterion(outputs, labels)\n",
        "\n",
        "                _, preds = torch.max(outputs.data, 1)\n",
        "                val_correct += (preds == labels).sum().item()\n",
        "                val_loss += loss.item()\n",
        "        print(f'Epoch {epoch + 1} validation accuracy: {val_correct/len(testing_data)*100:.2f}% validation loss: {val_loss/len(test_loader):.5f}')\n",
        "        val_loss_history.append(val_loss)\n",
        "\n",
        "    plt.plot(train_loss_history, label=\"Training Loss\")\n",
        "    plt.plot(val_loss_history, label=\"Validation Loss\")\n",
        "    plt.legend()\n",
        "    plt.show()\n",
        "\n",
        "# Train the model\n",
        "train_network_classification(mlp, train_loader, test_loader)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "n8_jBJg--ULC"
      },
      "source": [
        "## Testing your model\n",
        "Using the previously created `DataLoader` for the test set, compute the percentage of correct predictions using the highest probability prediction.\n",
        "\n",
        "If your accuracy is over 90%, great work, but see if you can push a bit further!\n",
        "If your accuracy is under 90%, you'll need to make improvements.\n",
        "Go back and check your model architecture, loss function, and optimizer to make sure they're appropriate for an image classification task."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "RWwh97TS-ULD"
      },
      "outputs": [],
      "source": [
        "## YOUR CODE HERE ##\n",
        "def evaluate_model(net, test_loader):\n",
        "    net.eval()\n",
        "    correct = 0\n",
        "    total = 0\n",
        "    with torch.no_grad():\n",
        "        for data in test_loader:\n",
        "            images, labels = data\n",
        "            if torch.cuda.is_available():\n",
        "                images, labels = images.cuda(), labels.cuda()\n",
        "            outputs = net(images)\n",
        "            _, predicted = torch.max(outputs.data, 1)\n",
        "            total += labels.size(0)\n",
        "            correct += (predicted == labels).sum().item()\n",
        "    accuracy = (correct / total) * 100\n",
        "    return accuracy\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rr4x1nl0jjWf",
        "outputId": "765696a9-86b4-4e08-a943-ae5fc3e7e812"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Test Accuracy: 97.46%\n"
          ]
        }
      ],
      "source": [
        "test_accuracy = evaluate_model(mlp, test_loader)\n",
        "print(f'Test Accuracy: {test_accuracy:.2f}%')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "P2rGTXkd-ULD"
      },
      "source": [
        "## Improving your model\n",
        "\n",
        "Once your model is done training, try tweaking your hyperparameters and training again below to improve your accuracy on the test set!"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 48,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 781
        },
        "id": "i17pqA-F-ULD",
        "outputId": "a5214c90-bc8d-42f7-f0b1-fd7bbd38397b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1 training accuracy: 98.87% training loss: 0.03656\n",
            "Epoch 1 validation accuracy: 97.41% validation loss: 0.12866\n",
            "Epoch 2 training accuracy: 98.94% training loss: 0.03234\n",
            "Epoch 2 validation accuracy: 97.80% validation loss: 0.12015\n",
            "Epoch 3 training accuracy: 99.06% training loss: 0.02899\n",
            "Epoch 3 validation accuracy: 97.60% validation loss: 0.12501\n",
            "Epoch 4 training accuracy: 99.10% training loss: 0.02828\n",
            "Epoch 4 validation accuracy: 97.51% validation loss: 0.12937\n",
            "Epoch 5 training accuracy: 99.03% training loss: 0.03064\n",
            "Epoch 5 validation accuracy: 97.07% validation loss: 0.14922\n",
            "Epoch 6 training accuracy: 99.09% training loss: 0.02878\n",
            "Epoch 6 validation accuracy: 97.24% validation loss: 0.15702\n",
            "Epoch 7 training accuracy: 99.13% training loss: 0.02813\n",
            "Epoch 7 validation accuracy: 97.59% validation loss: 0.12112\n",
            "Epoch 8 training accuracy: 99.12% training loss: 0.02853\n",
            "Epoch 8 validation accuracy: 97.60% validation loss: 0.12797\n",
            "Epoch 9 training accuracy: 99.15% training loss: 0.02685\n",
            "Epoch 9 validation accuracy: 97.70% validation loss: 0.12141\n",
            "Epoch 10 training accuracy: 99.18% training loss: 0.02466\n",
            "Epoch 10 validation accuracy: 97.65% validation loss: 0.13379\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "## YOUR CODE HERE ##\n",
        "def train_network_classification(net, train_loader, test_loader):\n",
        "    num_epochs = 10\n",
        "\n",
        "    criterion = nn.CrossEntropyLoss()\n",
        "    optimizer = optim.Adam(net.parameters(), lr=0.001)\n",
        "\n",
        "    train_loss_history = []\n",
        "    val_loss_history = []\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        net.train()\n",
        "        train_loss = 0.0\n",
        "        train_correct = 0\n",
        "        for i, data in enumerate(train_loader):\n",
        "            inputs, labels = data\n",
        "\n",
        "            if torch.cuda.is_available():\n",
        "                inputs, labels = inputs.cuda(), labels.cuda()\n",
        "\n",
        "            optimizer.zero_grad()\n",
        "            outputs = net(inputs)\n",
        "            loss = criterion(outputs, labels)\n",
        "            loss.backward()\n",
        "            optimizer.step()\n",
        "\n",
        "            _, preds = torch.max(outputs.data, 1)\n",
        "            train_correct += (preds == labels).sum().item()\n",
        "            train_loss += loss.item()\n",
        "        print(f'Epoch {epoch + 1} training accuracy: {train_correct/len(training_data)*100:.2f}% training loss: {train_loss/len(train_loader):.5f}')\n",
        "        train_loss_history.append(train_loss)\n",
        "\n",
        "        val_loss = 0.0\n",
        "        val_correct = 0\n",
        "        net.eval()\n",
        "        with torch.no_grad():\n",
        "            for inputs, labels in test_loader:\n",
        "                if torch.cuda.is_available():\n",
        "                    inputs, labels = inputs.cuda(), labels.cuda()\n",
        "\n",
        "                outputs = net(inputs)\n",
        "                loss = criterion(outputs, labels)\n",
        "\n",
        "                _, preds = torch.max(outputs.data, 1)\n",
        "                val_correct += (preds == labels).sum().item()\n",
        "                val_loss += loss.item()\n",
        "        print(f'Epoch {epoch + 1} validation accuracy: {val_correct/len(testing_data)*100:.2f}% validation loss: {val_loss/len(test_loader):.5f}')\n",
        "        val_loss_history.append(val_loss)\n",
        "\n",
        "    plt.plot(train_loss_history, label=\"Training Loss\")\n",
        "    plt.plot(val_loss_history, label=\"Validation Loss\")\n",
        "    plt.legend()\n",
        "    plt.show()\n",
        "\n",
        "# Train the model\n",
        "train_network_classification(mlp, train_loader, test_loader)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 52,
      "metadata": {
        "id": "0U43hb4_kcIS"
      },
      "outputs": [],
      "source": [
        "## YOUR CODE HERE ##\n",
        "def evaluate_model(net, test_loader):\n",
        "    net.eval()\n",
        "    correct = 0\n",
        "    total = 0\n",
        "    with torch.no_grad():\n",
        "        for data in test_loader:\n",
        "            images, labels = data\n",
        "            if torch.cuda.is_available():\n",
        "                images, labels = images.cuda(), labels.cuda()\n",
        "            outputs = net(images)\n",
        "            _, predicted = torch.max(outputs.data, 1)\n",
        "            total += labels.size(0)\n",
        "            correct += (predicted == labels).sum().item()\n",
        "    accuracy = (correct / total) * 100\n",
        "    return accuracy"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 53,
      "metadata": {
        "id": "tTkHj3PpkiRW",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "51ff02c7-a5aa-4a63-83ae-0bb72bcd5e3d"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Test Accuracy: 98.43%\n"
          ]
        }
      ],
      "source": [
        "test_accuracy = evaluate_model(mlp, test_loader)\n",
        "print(f'Test Accuracy: {test_accuracy:.2f}%')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GT50D3jt-ULE"
      },
      "source": [
        "## Saving your model\n",
        "Using `torch.save`, save your model for future loading."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 57,
      "metadata": {
        "id": "MtCykgP4-ULE"
      },
      "outputs": [],
      "source": [
        "## YOUR CODE HERE ##\n",
        "torch.save(mlp.state_dict(), 'mnist_mlp.pth')"
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "mvVN_Rb7opHY"
      },
      "execution_count": 56,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "ENJfwA6Zo8CT"
      },
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.6"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}